Earth and Planetary Science Letters 323-324 (2012) 60–71

Contents lists available at SciVerse ScienceDirect

Earth and Planetary Science Letters

j ourna l homepage: www.e lsev ie r .com/ locate /eps l

Asthenospheric flow and lithospheric evolution near the Mendocino Triple Junction

Kaijian Liu a,b,⁎, Alan Levander a, Yongbo Zhai a, Robert W. Porritt c, Richard M. Allen c

a Department of Earth Science, MS-126, Rice University, Houston, TX 77005, USAb Applied Physics Program, MS-366, Rice University, Houston, TX 77005, USAc Department of Earth and Planetary Sciences, University of California, Berkeley, CA 94720, USA

⁎ Corresponding author at: Dept. of Earth Science,77005, USA. Tel.: +1 713 348 5169.

E-mail address: lkjcammit@gmail.com (K. Liu).

0012-821X/$ – see front matter © 2012 Elsevier B.V. Aldoi:10.1016/j.epsl.2012.01.020

a b s t r a c t

a r t i c l e i n f o

Article history:Received 20 August 2011Received in revised form 16 January 2012Accepted 18 January 2012Available online xxxx

Editor: P. Shearer

Keywords:Mendocino Triple Junctionjoint inversionyoung asthenosphereslab windowmantle wedge

The migration of the Mendocino Triple Junction in northern California creates a complicated lithosphere–as-thenosphere boundary system at shallow depths (b60 km), following the progressive transition from Casca-dia subduction to San Andreas transform motion. It provides a natural laboratory to examine lithosphericevolution associated with slab removal in an active tectonic setting. However, pathways of the asthenospher-ic sources that fill the slab-free region created by the triple junction migration remain unclear. Previous activesource profiles and body-wave tomography are limited to either crustal or a deeper upper mantle scale, re-spectively. In this study, we present a three-dimensional shear velocity model from joint inversion of ambi-ent noise (8–30 s) and ballistic Rayleigh wave dispersion curves (22–100 s), as well as Ps receiver functions,using 111 stations from the Flexible Array Mendocino Experiment, USArray Transportable Array and the re-gional Berkeley Digital Seismic Network. In the crust, we have observed the low-Vs Franciscan Complex inthe Coast Ranges and the relatively high-Vs Great Valley ophiolite abutting the low-Vs Sierran basement.The low-Vs uppermost mantle imaged near the sudden steepening of the subducted oceanic slab extendsseismic evidence for forearc mantle serpentinization further south along the Cascadia margin. In additionto the asthenosphere beneath the Gorda plate, the joint inversion Vs model further identifies three otheryoung asthenospheres resulting from different partial melting mechanisms. Northward motion of the triplejunction causes asthenospheric flow both from under the Gorda plate and from the cooling former mantlewedge left under the Great Valley and Sierra Nevada, imaged from the joint inversion as a relatively deep(>75 km) low-Vs anomaly. These two mantle flows appear to begin mixing ~100 km south of the southernedge of the Gorda plate in the slab window region. We speculate that the latter provides the wedge-type geo-chemical signature seen in the Coast Range volcanic rocks, reconciling slab windowmodels and volcanic geo-chemistry. This ‘staggered’ upwelling model proposed here also explains the ~3 Myr delay in onset ofvolcanism after triple junction migration.

© 2012 Elsevier B.V. All rights reserved.

1. Introduction

The Mendocino Triple Junction (MTJ) is a transform–transform-trench triple junction forming the intersection of the Pacific, NorthAmerican and Gorda plates in northern California (e.g., Dickinson andSnyder, 1979; Furlong and Schwartz, 2004) (Fig. 1). The MTJ wasformed ~25–29 Ma by the impingement of the Pacific–Farallon ridgeon the North American trench, initiating San Andreas fault (SAF)strike-slip motion (Atwater, 1970). North of the MTJ, the subductionof the young Gorda plate beneath North America forms the southernterminus of the Cascadia subduction zone. The northwestward migra-tion of the MTJ, resulting in the progressive replacement of the subduc-tion zone by the transform margin, has profoundly altered thelithospheric structure of the western North American plate boundary

Rice University, Houston, TX

l rights reserved.

through a variety of tectonic and magmatic processes (e.g., Dickinson,1997). The triple junction region has two distinct volcanic systems:the southernmost Cascade Quaternary volcanic centers (e.g. MountShasta and Lassen Peak), resulting from mantle wedge melting abovethe Gorda plate (Leeman et al., 2005) and the northern Coast Range vol-canic system, consisting of isolated volcanic centerswith northward de-creasing ages aligned along the strike of the range, the youngest beingthe Clear Lake volcanic fields (~2–0.1 Ma) (Dickinson, 1997; Furlongand Schwartz, 2004; Johnson and O'Neil, 1984).

The northwestward migrating MTJ leaves in its wake a lithospher-ic gap at the southern edge of the Gorda plate (SEDGE) (Dickinsonand Snyder, 1979; Furlong and Schwartz, 2004; Levander et al.,1998; Liu and Furlong, 1992), thought to be filled by asthenosphericflow, the subject of this paper. Late-Cenozoic Coast Range volcanismand toroidal mantle flow are attributed to flow into the gap (Eakinet al., 2010; Zandt and Humphreys, 2008). Evidence supporting thismodel includes surface heat flow (Lachenbruch and Sass, 1980), pat-terns of volcanism (e.g., Dickinson, 1997), active and passive source

http://dx.doi.org/10.1016/j.epsl.2012.01.020mailto:lkjcammit@gmail.comhttp://dx.doi.org/10.1016/j.epsl.2012.01.020http://www.sciencedirect.com/science/journal/0012821X

−128˚ −126˚ −124˚ −122˚ −120˚ −118˚

38˚

40˚

42˚

N

GordaPlate

PacificPlate

Coast R

angesSan Andreas Fault

Cascad

ia Su

bd

uctio

n Z

on

e

San Francisco

GreatValley

SierraNevada

Clear Lake

−6 −4 −2 0 2 4

MTJ

km

North America

Klam

ath Mts. Basin & Range

FAMEBDSNUSArray/TA

Fig. 1. Study region (yellow box) with topography and principal tectonic features. We analyzed broadband data from 75 FAME stations (red triangles), 22 USArray/TA stations (darkinverse triangles), and 14 Berkeley Digital Seismic Network (BDSN) stations (blue diamonds). Volcanoes are plotted as white dots, including Clear Lake, in the Coast Ranges, and Mt.Shasta and Lassen Peak in the Cascade Ranges. Dark dashed line shows the SAF along the Pacific-North American plate boundary. Purple lines mark the major tectonic boundaries.Orange arrows indicate Pacific and Gorda plate motions relative to North America. (For interpretation of the references to color in this figure legend, the reader is referred to theweb version of this article.)

61K. Liu et al. / Earth and Planetary Science Letters 323-324 (2012) 60–71

seismic images (Beaudoin et al., 1996; Beaudoin et al., 1998; Benz etal., 1992; Levander et al., 1998; Zhai, 2010), and geodynamicmodeling(Liu and Furlong, 1992; Zandt and Furlong, 1982). The origin of theslab window asthenosphere and the flow paths into the gap followingslab removal are debated (e.g., Cole and Basu, 1995; Hole et al., 1998;Schmitt et al., 2006; Whitlock et al., 2001). Previous asthenosphericupwelling models fall into two categories, involving either subconti-nental (e.g., Johnson and O'Neil, 1984; Liu and Furlong, 1992;Schmitt et al., 2006; Whitlock et al., 2001) or suboceanic (e.g., Coleand Basu, 1995; Zandt and Humphreys, 2008; Zhai, 2010) sources.

Seismic converted wave imaging, tomography, and SKS split timescan be used to identify the slab window structure, asthenospheric seis-mic velocities, and the directions of mantle flow in the MTJ region. Inthis study, we constructed a three-dimensional shear velocity modelby the joint inversion of Ps receiver functions and fundamental-modeRayleigh wave dispersion data (8–100 s) from the EarthScope flexibleand transportable arrays (Fig. 1). The joint inversion provides bettervertical resolution of upper mantle structures in the resolution gapthat exists from roughly the Moho to the asthenosphere between tele-seismic body-wave tomography (Benz et al., 1992; Obrebski et al.,2010; Schmandt and Humphreys, 2010) and active seismic studies(Beaudoin et al., 1996, 1998; Henstock et al., 1997; Levander et al.,1998).

2. FAME and USArray datasets

We used data from the Flexible Array Mendocino Experiment(FAME) network, which consists of 75 broadband seismic stationsdeployed in northern California from July 2007 to June 2009. Addition-ally, we incorporated data from 22 USArray Transportable Array (TA)stations and 14 Berkeley Digital Seismic Network (BDSN) stations, asthey have temporal overlap with the FAME experiment (Fig. 1). Usinga total of 111 stations, we first measured fundamental-mode Rayleighwave phase velocities (22–100 s). The resulting phase velocities werecombined with short-periods (8–30 s) from an ambient noise

tomography (ANT) study (Porritt et al., 2011), yielding broader bandsurface wave dispersion curves (8–100 s). We inverted for the isotropicVs structure using the joint inversion of the combined dispersion curvesand the Ps receiver functions (Zhai, 2010) made with data from thesame stations.

2.1. Ballistic Rayleigh wave phase velocity (22–100 s)

We use 224 teleseismic earthquakes with epicentral distance be-tween 30° and 120°, magnitude >5.5 and depths b70 km (Fig. 2a,b). To account for the nonplanar energy in the fundamental-modeRayleigh wavefield caused by multipathing or small-scale scattering,we applied the modified two-plane wave technique (Forsyth and Li,2005; Liu et al., 2011; Yang and Forsyth, 2006), which incorporatesfinite-frequency sensitivity kernels for both amplitude and phase(Yang and Forsyth, 2006). Here, the recorded wavefield is repre-sented approximately by the sum of two interfering plane waves.We inverted the Rayleigh wave phase velocities for 13 frequencybands from 22 to 100 s based on a grid of 0.25°×0.25° near the MTJ re-gion. The combination of the FAME, USArray/TA and the BerkeleyBDSN stations provides dense enough coverage to resolve the rela-tively short wavelength phase velocity variations.

2.2. Ambient noise dispersion data (8–30 s)

Ambient noise tomography computes the cross-correlation wave-form of seismic noise between station pairs as an estimate of theGreen's functions to extract the inter-station phase (e.g., Bensen etal., 2007; Shapiro et al., 2005; Snieder, 2004; Yao et al., 2006). TheANT provides shorter-period dispersion data (8–30 s) than theballistic earthquake tomography described above, which greatlyincreases the study's sensitivity to crustal shear velocity (Porritt etal., 2011). In the overlapping period range (22–30 s), there is goodconsistency between the ballistic and ANT phase velocities (Fig. 3c).

22 25 27 30 34 40 45 50 59 67 77 87 1000

2000

4000

6000

8000

10000

12000

14000

Period (s)

Num

ber

of r

aypa

ths

a)

b)

Fig. 2. (a) Azimuthal and distance distribution of the 224 teleseismic events (red dots)for the finite-frequency Rayleigh wave tomography. (b). Number of raypaths for eachperiod (frequency) band from 22 to 100 s in the surface wave study. (For interpretationof the references to color in this figure legend, the reader is referred to the web versionof this article.)

62 K. Liu et al. / Earth and Planetary Science Letters 323-324 (2012) 60–71

2.3. Ps receiver function

We include the Ps receiver functions (1 and 2 Hz), generated usingFAME data (Zhai, 2010), in a joint phase velocity/receiver function in-version to provide a stronger constraint on seismic discontinuitystructure in the crust and upper mantle. The station-gather receiverfunctions were made from 186 teleseismic events (35°≤Δ≤90°,Mw>5.7) using time-domain iterative deconvolution technique(Ligorria and Ammon, 1999). In our joint inversion, we selected twooverlapping frequency band receiver functions with hi-cut frequen-cies of 1 and 2 Hz (corresponding Gaussian widths of 2.07 and 4.14,respectively). The receiver functions increase sensitivity to the under-lying seismic discontinuity structure, while use of two different fre-quencies improves our ability to differentiate between first orderdiscontinuities and Vs gradients, which both cause P-to-S conversions(e.g., Julia et al., 2000). To enhance the signal-to-noise ratio, we divid-ed the receiver functions under each station into three groups withdifferent ray parameters (b0.05 s/km; 0.05–0.06 s/km; >0.06 s/km).After the Ps time move-out correction (e.g., Rondenay, 2009) was ap-plied and assuming the 1-D background Tectonic North America(TNA) shear model (Grand and Helmberger, 1984), we stacked thethree groups of receiver functions individually at the central ray pa-rameters of 0.045, 0.055 and 0.065 s/km, as well as collectively at

0.05 s/km (e.g., Fig. 3a). This stacking technique reduces sensitivityto deep discontinuities and crustal reverberations while enhancingthe shallow conversion signals; therefore, we used short time win-dows from −5 to 10 s (equivalent to about 100 km depth) for bothlow and high frequency receiver functions (Fig. 3a and S1b). Theshort windows prevent modeling multiple reflections in the inver-sion. In this region crustal thickness varies rapidly laterally and verti-cally in both coast normal and coast parallel directions. Crustalmultiples are poorly predicted by simple 1-D models in the presenceof rapid crustal thickness changes. The significant signals for ourstudy, the Moho and lithosphere–asthenosphere boundary (LAB),are contained within the first 10 s after P, since LAB depths in this re-gion are typically b100 km (Li et al., 2007; Zhai, 2010).

3. Joint inversion method

To reduce the nonuniqueness of the Vs inversion from each singledataset, we jointly invert the shear-velocity structure beneath eachstation from the three independent observations (e.g., Fig. 3): (1)22–100 s phase velocity from finite-frequency Rayleigh wave tomog-raphy, (2) 8–30 s phase velocity from ambient noise tomography(Porritt et al., 2011), and (3) Ps receiver functions at low (1 Hz) andhigh (2 Hz) frequency bands (Zhai, 2010). The dispersion curvesfrom the first two observations provide complementary constrainton the average absolute Vs from the shallow crust to the upper man-tle, while the Ps receiver functions are primarily sensitive to the shearimpedance variations with depth. The combined constraints signifi-cantly decrease the nonuniqueness of the Vs inversion, and thus im-prove the model reliability for the crust and upper mantle (Fig. 4).The joint inversion also helps refine Moho and LAB depth estimatesin regions of ambiguity in the receiver function images (Fig. 5a, b).

We used the Computer Programs in Seismology package (Herrmannand Ammon, 2002) to invert the 1-D shear velocity models beneath the111 stations (Fig. 3). We modified the cost function here to include allthree datasets with assigned weight for the joint inversion:

1−pNrf Npts

XNrfi¼1

XNptsj¼1

‖εrfij ‖2 þ p

NdspwRWT

XNRWTi¼1

‖εRWTi ‖2 þwANT

XNANTi¼1

‖εANTi ‖2

( )ð1Þ

where εi≡(diOBS−diPRE)/σi is the relative residual between the ob-served and synthetic data, Nrf, Ndsp, NRWT, NANT are the total numbersof receiver functions, periods at which dispersion was measured,Rayleigh wave ballistic phase velocities, and ANT phase velocities, re-spectively. Npts is the total number of sampling points in each receiverfunction, and p is the influence factor controlling the constrainingweight given to the receiver function data and the dispersion data(Fig. S1a). In order to avoid overemphasis of the ambient noise datadue to closer period sampling, we assigned different weights to thedispersion data from earthquake tomography and ambient noiseanalysis in the form of wi=Ndsp/2Ni(i=RWT,ANT). Thirty iterationswere used for the nonlinear inversion for the Vs model startingfrom the TNA reference model with no a priori crustal structure(Fig. 3b). A slight increase of the weight term at the Moho depth esti-mated from the receiver functions allowed the possibility of large ve-locity variations under smoothing regularization. If the a priori Mohodepth differs greatly from the true model, the iterative inversion willsearch for the true depth. The joint inversion also helps to improvethe Moho and LAB depth estimates with respect to those made fromreceiver functions alone (Fig. 5).

After completion of the station-by-station joint inversion for 1-DVs structure, the 1-D profiles were collected into a 3-D volume and in-terpolated onto a regular horizontal and vertical grid. Although theinterpolation smoothes out sharp velocity changes, the lateral resolu-tion in the 3-D volume is as good as the station spacing (~35 km) al-lows. The horizontal resolution is controlled by a number of factors,

-3 0 3 6 9

2.07 9 6 . 8 0 0 . 0 4 5

2.07 9 8 . 9 6 0 . 0 5 0

2.07 9 5 . 3 3 0 . 0 5 5

2.07 9 7 . 7 6 0 . 0 6 5

4.14 9 5 . 7 3 0 . 0 4 5

4.14 9 8 . 6 2 0 . 0 5 0

4.14 9 6 . 0 3 0 . 0 5 5

4.14 9 5 . 0 4 0 . 0 6 5

3.00 4.20

120

90

60

30

0

Dep

th (

km)

0 20 40 60 80 100 1202.6

2.8

3

3.2

3.4

3.6

3.8

4

4.2

ANTBallisticPredicted

Period (s)Time (s)P

hase

vel

ocity

(km

/s)

Current modelStarting 1-D model

c) Rayleigh wave dispersion curvesb) Vs (km/s)a) Ps receiver functions

Fig. 3. One example of a Rayleigh wave-receiver function joint inversion at station XQ.ME39. (a) We use 1 and 2 Hz Ps receiver functions binned at 0.045, 0.050, 0.055 and 0.065 s/km. Red and blue are the predicted and observed receiver functions, respectively. Top four traces are the 1 Hz receiver functions, and bottom four are the 2 Hz receiver functions.Each trace includes Gaussian filter parameters on the left, and percentage of model fit and ray parameter on the right. (b) The starting Vs model for the joint inversion is a constantvelocity using the uppermost mantle velocity from the Tectonic North America (TNA) model (dashed blue line) with no a priori information for the crustal thickness and velocities.Red line is the final Vs model after 30 iterations. (c) Both ANT (dark triangles) and ballistic (blue dots) Rayleigh wave phase velocities are used in the joint inversion, while thepredicted dispersion curve (red) is calculated from the final Vs model from (b). (For interpretation of the references to color in this figure legend, the reader is referred to theweb version of this article.)

38˚

40˚

42˚

2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4a) Crust/Uppermost mantle Vs

10 km 20 km 26 km 32 km

−124˚ −122˚ −120˚38˚

40˚

42˚

2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6b) Mantle Vs

50 km

−124˚ −122˚ −120˚

60 km

−124˚ −122˚ −120˚

80 km

−124˚ −122˚ −120˚

100 km

km/s

km/s

Gorda asthenosphere

GV Ophiolite?GV Ophiolite?

Franciscan

Complex

Franciscan

Complex

KlamathKlamath

Slab window

SEDGE SEDGE

Slab window

arc volcanoes

GV-SN source GV-SN source

S1

S4

S3

S2

Fig. 4. Shear velocity maps at depths from 10 to 100 km from the joint inversion of ambient noise, ballistic Rayleigh wave dispersion and Ps receiver functions. Red triangles are thevolcanic fields. Dark dashed line is the San Andreas Fault. The red dashed line is the inferred SEDGE. Note that the crust/uppermost mantle (a) and upper mantle (b) velocity mapsare plotted with different color scales.

63K. Liu et al. / Earth and Planetary Science Letters 323-324 (2012) 60–71

a) Moho depth (km)

−124˚ −122˚ −120˚38˚

40˚

42˚

20 25 30 35 40 45 50

b) LAB depth (km)

−124˚ −122˚ −120˚38˚

40˚

42˚

40 50 60 70 80 90 100

GordaMantleWedge

SlabWindow

‘GV-SN’ anomaly

Subductionregime

Transformregime

Arc moho

Serpentinizedzone

Moho‘hole’

SEDGE

SEDGE

A’

D’

C

D

B’B

A

C’

1

2

65

43

Fig. 5. Estimated (a) Moho and (b) LAB depth maps. Blue dashed line marks the SEDGE.The relative thick forearc crust refers to the oceanic Moho. The white dashed line in(a) outlines the serpentinized zone, where we observe low-Vs uppermost mantle.Dark thin lines (AA′, BB′, CC′ and DD′) in (b) show the locations of cross-sectionsshown in Fig. 6 and 7. The numbers 1–6 identify the locations of the Vs-z profiles inFig. S4. (For interpretation of the references to color in this figure legend, the readeris referred to the web version of this article.)

64 K. Liu et al. / Earth and Planetary Science Letters 323-324 (2012) 60–71

including the wavelengths of the surface waves used in the tomo-graphy, the assumption that all the signals can be described by two-plane waves (resolution estimated as inter-station spacing of~35 km for periods b40 s) (Yang et al., 2008), and the 1-D inversion.Vertical resolution is controlled by a combination of the Rayleighwave's vertical averaging of velocity, the receiver functions' abilityto identify sharp gradients, and the weight given to each in the inver-sion (Fig. S1a). For the dispersion data, the constraint on the absoluteVs at different periods decays with depth, with the sensitivity peakslocated at depths of approximately a third of a wavelength (e.g., Liuet al., 2011). The receiver function's vertical resolution is no betterthan ¼ of a wavelength; therefore, for a frequency of 1 Hz, a shear

wave at the Moho would be unable to distinguish between an impe-dance (shear velocity times density) gradient occurring over about1 km and a step in impedance.

4. Result and discussion

4.1. Crustal heterogeneity: Franciscan Complex, ophiolite, and crustalmelt

The crustal heterogeneity revealed by the joint inversion Vs modelcorrelates well with surface geological features and reflects the com-plicated tectonic history in the accretionary wedge near the continen-tal margin. The upper crustal low-Vs values (b3.0 km/s) in the CoastRanges (Fig. 4a) are Franciscan Complex accretionary wedge, low tohigh-grade metamorphic rocks (e.g., Blake et al., 1985). Along theCoast Range Fault (10 and 20 km depth, Fig. 4a), the Franciscan Com-plex is juxtaposed against the Klamath block, which has higher veloc-ities (3.4–3.6 km/s), reflecting the Trinity ultramafic sheet and NorthFork ophiolite (Thurber et al., 2009; Zucca et al., 1986). The low-velocity layer underlying the Klamath block is most likely accretion-ary wedge sedimentary rocks, or possibly peridotite serpentinizedunder low-temperature conditions. In the Great Valley, the shallowlow-Vs layer is an average of the thick sedimentary basin and the un-derlying crust, since we do not account directly for the basin in thestarting model. The underlying high-velocity body in the valley (20and 26 km depth, Fig. 4a) is interpreted as the Great Valley ophiolite,probably resulting from back arc obduction of oceanic crust/mantleover the continental material during the Jurassic orogeny (Godfreyand Klemperer, 1998). The ophiolite under the Great Valley can bedistinguished from the lower Vs Sierran basement which is eitherthe Foothills Metamorphic Complex and/or the intrusive batholith(Godfrey and Klemperer, 1998; Godfrey et al., 1997). Under the Cas-cadia volcanoes such as Lassen Peak and Mt. Shasta, the pronouncedlow-velocity layer (Vs~3.0 km/s) in the crust (e.g. Fig. 4a) suggestscrustal partial melt above the mantle wedge.

4.2. Comparison with MTJSE Line 9

The MTJ marks the transition from a subduction (convergent) re-gime to a transform regime resulting from the northward migrationof the Pacific and Gorda plates relative to North America, which pro-foundly influenced North American lithospheric structure (Beaudoinet al., 1996, 1998; Dickinson and Snyder, 1979; Furlong andSchwartz, 2004; Henstock et al., 1997; Levander et al., 1998). Wecompare our joint inversion Vs cross-section with the correspondingreceiver function image (Zhai, 2010) and the crustal reflection/refrac-tion P-wave velocity models (Beaudoin et al., 1996, 1998), along Line9 from the MTJ Seismic Experiment (MTJSE) (Fig. 6a–c).

4.3. Gorda plate

At the northern end of the profile, the high-Vs body (3.6–4.5 km/s;Fig. 6c) is the subducting Gorda plate. The SEDGE is reasonably welldefined as the limit of high velocities to the south, marking the youn-gest part of the slab-free region in the transform regime nearx=90 km (Fig. 6). The high Vs region indicative of the plate is bound-ed above by a positive receiver function event, and below by a nega-tive event. The former, a broad pulse, we interpret as interferingsignals from the oceanic Moho and top of the oceanic crust(~20–25 km). The top of the oceanic crust in the Vs image is consis-tent with strong reflections from the active source study (Fig. 6a).The Gorda oceanic crust appears as intermediate velocities(~3.9–4.1 km/s) between the reflections from the top of the crustand the Moho (dark and white lines in Fig. 6b, respectively). The neg-ative receiver function signal at the bottom of the high velocity plate(~4.4–4.5 km/s) we interpret as the Gorda LAB at ~70 km. The

0

25

50

75

100

125

Dep

th z

(km

)

0 50 100 150 200

Distance x (km)

CCP RF −0.5

0.0

0.5

Amplitude

0

25

50

75

100

125

Dep

th z

(km

)

0 50 100 150 200

2.8

3.2

3.6

4.0

4.4

Vs (km/s)

Joint Inv. Vs

-505

10152025303540

Dep

th (

km)

0 50 100 150 200 250Distance (km)

2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5Vp (km/s)

SP 901 902 903 904 905 906 907 908 909 910 911

MTJSE Line 9Line 1 NWSE

LAB Gorda Plate

Slab Window SEDGE

Top OC

Gorda Moho

TransformMoho

Multiple

Subduction regimeTransform regime

CLV

VBasaltIntrusionsMFZ

Distance x (km)

a)

b)

c)

Gorda slabSEDGESlab

WindowLAB

Subduction regimeTransform regime

CLV

Gorda asthenosphere

A

A A’

A’

OC

N.A. crust

Lake Pillsbury

Fig. 6. Comparison of cross-sections from the joint inversion Vs model with other seismic results along the MTJ Seismic Experiment (MTJSE) Line 9 (A-A′ in Fig. 5b). (a) A crustal Vpmodel similar to that of Beaudoin et al. (1996), (b) Ps receiver function image (Zhai, 2010), and (c) the joint inversion Vs model. Dark dots in (b) are earthquake hypocenters. In thejoint inversion Vs image (c), we superimpose the top of the oceanic crust (OC) (white solid line), Gorda Moho (dark solid line), and the LAB depth under the Gorda plate (whitedashed line) determined from the active source model and FAME receiver functions. Purple dashed line is the estimated LAB from the joint inversion Vs model taken as the center ofthe negative Vs gradient. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

65K. Liu et al. / Earth and Planetary Science Letters 323-324 (2012) 60–71

heterogeneous velocity structure within the Gorda lithosphere coin-cides with the location of the strong internal deformation as indicatedby seismicity and seafloor fault offsets (Gulick et al., 2001; Wilson,

1986, 1989). The Gorda plate dips southward toward the MTJ, termi-nating under the thickest accretionary wedge, also at x=~90 km(Fig. 6c). It has been hypothesized that mechanical coupling between

66 K. Liu et al. / Earth and Planetary Science Letters 323-324 (2012) 60–71

the slab and the overlying crust cause an ephemeral crustal root toform near the SEDGE (Figs. 6c and S4) (Beaudoin et al., 1998;Furlong et al., 2003).

North of the MTJ, we clearly image the subducting Gorda plate(high-Vs body in Figs. 4b, 6c, 7a) beneath North America. The forearccrust thickens eastward from ~25 to ~40 km, with thickening termi-nating at the location of rapid plate steepening from ~10° to ~25°dip (Fig. 7a). The estimated plate thickness (~30–40 km) from the re-ceiver function and joint inversion Vs model agrees well with the

a)

b)0

25

50

75

100

125

Dep

th (

km)

39.5 40.0 40Latitude

0

25

50

75

100

125

Dep

th (

km)

−124 −123Longitude

lat=41.0

0

25

50

75

100

125

Dep

th (

km)

−123 −122Longitude

La

‘slab window’

‘GV-

lat=39.0

Subducted Gorda

Serpentin

Gorda asthenosphere

M

GV-SN asthenosphere

LAB

B

D

C

c)

Fig. 7. Vs cross-section along lines B-B′, C-C′, and D-D′ in Fig. 5b. (a) Shows the subductedvolcanoes, the low Vs mantle wedge lies above a high-Vs body we interpret as the subductto the mantle wedge asthenosphere. (c) Shows the asthenospheric flow from the ‘GV-SN’ andistribution.

thickness predicted from the half-space cooling model for oceaniclithosphere (~29–36 km), using ~8–13 Ma as the age of subductedGorda plate, and 1100 °C as the temperature at the base of the ther-mal boundary layer and 1300 °C as the (convecting) ambient mantle,respectively. West of the change in slab dip, we observe a shallow LAB(~50–60 km) at the base of the Gorda plate. Low-Vs (b4.2 km/s) inthe Gorda asthenosphere is likely due to proximity to the Gordaridge, and/or shear melt (Kawakatsu et al., 2009) at the base of thesubducted slab.

.5 41.0 41.5

−122 −121

−1212.8

3.0

3.2

3.4

3.6

3.8

4.0

4.2

4.4

4.6

Vs (km/s)

ssen Peak

SN’ asthenosphere

slab

Mantle Wedge

ized mantle

antle Wedge

Subducted Gorda slab

Mt. Shasta

D’

C’

B’

Dehydration

slab, serpentinized forearc mantle, and low Vs mantle wedge. (b) Under the Cascadiaed Gorda plate. The ‘GV-SN’ asthenosphere is relatively deep (>75 km) in comparisonomaly to the slab window from east to west. White dots in (a) and (c) are the seismicity

67K. Liu et al. / Earth and Planetary Science Letters 323-324 (2012) 60–71

4.4. Slab-free window

In the transform regime, we interpret the shallow low-velocitybody between x=0–80 km at 35–75 km depth as asthenosphere inthe slab-free region (Figs. 4b and 6c). North of about x=50 km, thesub-Moho shear velocity in the ‘slab-free window’ is b4.2–4.3 km/sbut velocity decreases (b4.0 km/s) moving southward betweenx=0–50 km. The lowest velocities are at ~35 km depth beneath theClear Lake volcanic field (Figs. 6c and S4). The continuity of the lowvelocity anomaly from beneath Gorda to the transform regimewould allow mantle flow of sub-Gorda asthenosphere into the slabwindow. The significant decrease in velocity accompanying this alsosuggests generation of a small fraction of partial melt.

In addition to the Clear Lake volcanic field, beneath the Lake Pills-bury region, the MTJSE active seismic data have imaged highly reflec-tive bodies in the lower crust, interpreted as partially molten basaltsills (Levander et al., 1998). The area has been the site of midcrustalseismicity interpreted as diking events (e.g., Furlong and Schwartz,2004). The LAB in the slab-gap region (at z~30–40 km) is 30–45 kmshallower than that beneath the Gorda plate (z~60–75 km), resultingin decompression melting of ascending asthenosphere from beneathGorda. Pressure-released basaltic melt can be generated at the baseof the transform crust (Henstock and Levander, 2000), causing theslab window magmatism at Lake Pillsbury and Clear Lake. However,the apparent thickening asthenosphere and thinning of the litho-sphere southward away from the MTJ are of the exact opposite thegeometry predicted by the simple slab window model as describedby Dickinson and Snyder (1979). We suggest that the asthenospherethickens to the south by upward and lateral flow of hydrated mantlewedge, left behind as the triple junction migrates, as we describe fur-ther below.

4.5. Forearc serpentinization

Near the northernmost Great Valley, we detect weak or absent PsMoho signals, and a complicated Vs structure at the base of the crust,which we term a ‘Moho hole’ (Figs. 5a and S3) (Zhai, 2010). Serpen-tinization of the forearc upper mantle has been widely inferred to ex-tend north along the Cascadia margin, where it is characterized bylow uppermost mantle Vs, weak or reversed Ps conversions(Bostock et al., 2002), weak or missing PmP/Pn signals (Beaudoin etal., 1996, 1998; Brocher et al., 2003), and characteristic magneticanomalies (Blakely et al., 2005). Using the Vs-serpentinization rela-tion (Bostock et al., 2002), the uppermost mantle Vs (~3.7 km/s at~40 km depth, Fig. 7a) gives an estimate of up to ~30–40% serpentini-zation of the northern California forearc, an intermediate value be-tween that suggested for Cascadia (Bostock et al., 2002) to thenorth and a more normal, apparently unserpentinized northernGreat Valley ophiolite (Godfrey and Klemperer, 1998) to the south.The transition of serpentinized-to-unserpentinized mantle near theSEDGE suggests further mantle dehydration in the post-subductionregime following slab removal (Fulton and Saffer, 2009), and recy-cling of the oceanic slab-derived components into the continentalmantle. At ~50–75 km, low velocities immediately east and abovethe subducting Gorda plate (Fig. 7a) suggest the possibility of ongoingdehydration of the plate (e.g., Hacker et al., 2003).

4.6. Young asthenospheres

To the south, the absence of the high-Vs slab defines the SEDGE,similar to body-wave tomography results (Obrebski et al., 2010). Be-sides the Gorda asthenosphere (S1 in Fig. 4b), we observe three addi-tional low-Vs regions (S2, S3, and S4 in Fig. 4b) in the upper mantle,and identify three modern shallow LABs (Fig. 5b), corresponding tothe Cascadia mantle wedge (S3), the San Andreas transform margin(S2), and another low-Vs region (S4), the ‘GV-SN’ anomaly that we

describe below. The SAF LAB shallows rapidly from the subduction re-gime to b50 km depth beneath Clear Lake. To the east of Clear Lake atgreater depths (>75 km) we find low velocities (b4.1 km/s) underthe Great Valley and the western Sierra Nevada block, which weterm the “GV-SN” anomaly. We interpret this feature as the coolingformer mantle wedge. The LAB at the top of the wedge is deeper(~75 km) than elsewhere, likely due to development of a conduc-tively cooling boundary layer and/or a remnant lithospheric lidunder North America.

4.6.1. Mantle wedge and Cascadia volcanoesWe interpret the shallow mantle wedge LAB under the Cascadia

arc as resulting from hydration-induced decompression melting,causing a significant drop in Vs (Leeman et al., 2005). The extremelylow velocities (~3.6 km/s at 60–80 km depth) in the wedge (Fig. 4band 7a, b) can be attributed to partial melts from dewatering of thesubducting oceanic lithosphere and convective upwelling (Leemanet al., 2005). Cross-sections at 41.0 N (Fig. 7) and 40.6 N (Fig. S2)show continuous low velocities from the top of the Gorda plate at50–75 km depth to directly beneath the Cascadia arc, suggestingthat these are fluid pathways from the dehydrating slab. Fluid-mediated decompression melting significantly shallows the LABdepths to 50–60 km, slightly subcrustal, and drives Cascadia magma-tism by allowing melt to infiltrate the base of the North Americanlithosphere. The mantle wedge magmatism causes the arc volcanoesof the southernmost Cascadia Range: Lassen Peak and Mount Shasta.We interpret the low Vs upper mantle in the back arc as a continua-tion of the mantle wedge asthenosphere.

4.6.2. ‘GV-SN’ anomalyAs the MTJ migrates northwestward, it leaves behind a mantle

wedge. We suggest that the ‘GV-SN’ anomaly is the former mantlewedge, located to the south, along strike with the current mantlewedge. However, we note that the LAB at ~75 km is deeper than theactive mantle wedge by 15–20 km. We interpret the southeastwardthickening high-Vs lithospheric lid (~10–30 km thick, Fig. 7b) as agrowing thermal boundary layer caused by cooling of former mantlewedge asthenosphere south of the SEDGE.

Slab removal changes the mantle flow field by removing the driv-ing force of mantle wedge counter flow. If the asthenosphere underthe Gorda plate can flow into the slab-free window, then Gorda as-thenosphere beneath deeper parts of the subducting slab can alsoflow into the former mantle wedge. It seems likely that the formermantle wedge can also flow laterally and upward into the slab win-dow. The GV-SN low velocity region is likely either, former mantlewedge, or former mantle wedge mixed with Gorda asthenosphere.

4.6.3. SAF transform regimeBeneath the transform regime, the significant decrease in Vs

(b4.0–4.2 km/s, Figs. 4b and 6c) and very shallow LAB (~30–40 km)south of SEDGE are attributed to partial melts produced by decom-pression melting of upwelling asthenosphere from beneath Gorda(Henstock and Levander, 2000). The melting mechanism here differsfrom that beneath the Gorda plate (shear lensing) and mantle wedge(hydration-induced reduction of the solidus). A certain amount of themelt is accreted at the base of the transform crust (Levander et al.,1998), creating some of the slab window magmatism in Clear Lakeand elsewhere in the Coast Ranges. The shallowest LAB under theCoast Ranges is between Clear Lake and Lake Pillsbury, the latterbeing predicted as a location for future volcanism (Levander et al.,1998). We interpret the continuity of low velocities (Fig. 6c) andSKS split patterns (Fig. 8) to indicate mantle flow into the CoastRanges slab window from beneath the Gorda plate and from theman-tle wedge.

1 second

elevation (km)

Subductionregion

‘SlabWindow’

‘GV-SN’ anomaly

SEDGE

Fig. 8. Fast polarization directions of SKS splits in theMTJ regionwith absolute platemotions (yellow arrows) of Gorda, Pacific and North American plates. Shear wave splitting results areplotted as different colors: dark fromEakin et al. (2010), blue fromFouch andWest (in preparation) and red from Liu (2009). Dark and orange dashed contours are the slabwindow regionin the transform regime and the “GV-SN” anomaly inferred from joint inversion in this study. Blue dashed line is the current SEDGE as shown in Fig. 5, while purple dashed line showsprevious SEDGE location. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

68 K. Liu et al. / Earth and Planetary Science Letters 323-324 (2012) 60–71

4.7. Mantle flow

Geochemical data from the northern Coast Range volcanics showstrong similarity to the mantle wedge, both having enriched large-ion lithophile and depleted high-field-strength elements (e.g.,Furlong and Schwartz, 2004; Whitlock et al., 2001). Hole et al. (Holeet al., 1998) argued that a mantle wedge-derived origin could betterfit the Coast Range heat flow data (Lachenbruch and Sass, 1980),i.e., mantle flow from the wedge to the shallow slab-free region. Incontrast, initially low 87Sr/86Sr ratios and high εNd of the volcanicrocks (Cole and Basu, 1995) in the Coast Ranges south of Clear Lakeimply that the slab window was first occupied by asthenospherefrom an oceanic source. This is consistent with the continuity of theLAB topography in the slab window south of the SEDGE (Zhai, 2010).

It has been suggested that a circular pattern of SKS splits observedin central California, Nevada, Utah and Oregon result from large scaletoroidal mantle flow emanating from beneath the Gorda plate alongthe SEDGE (e.g., Eakin et al., 2010; Zandt and Humphreys, 2008).The scale of this flow pattern is almost entirely out of our study re-gion. For example Fig. 1 of Piromallo et al. (2006) and Fig. 2b ofZandt and Humphreys (2008) suggest that immediately south of the

MTJ, mantle flow from under the sinking Gorda plate will be normalto the plate edge, meaning to the south-southwest in the Mendocinoregion, and of small magnitude. SKS splitting measurements fromEakin et al. (2010), Liu (2009) and Fouch and West (in preparation)show a nearly SEDGE parallel signature (Fig. 8) with relatively largemagnitudes (~1 s). The inconsistency between the observed SKSsplits in the MTJ region and the predictions of toroidal flow are the re-sult of the scale of observation. We suggest that locally the plate edgecauses a more complicated mantle flow field as the sinking platedrags viscous mantle down with it.

The low Vs in the GV-SN anomaly, the SKS splits, and the appear-ance of a mantle wedge signature in the geochemical data (Whitlocket al., 2001) lead us to hypothesize a ‘staggered’, two-stage upwellingmodel (Figs. 9 and 10). We suggest that as North America moves tothe northwest, the opening slab window is filled from two sources:The sub-Gorda asthenosphere flows upward and to the southeastinto the slab window, while simultaneously abandoned mantlewedge flows upward and to the southwest into the slab window(Figs. 9 and 10b, c). The abandoned wedge is left behind as the ‘GV-SN’ asthenosphere (Figs. 7c and 8), and is likely still contributing tothe slab window asthenosphere by upward and lateral flow. The

−124˚−123˚

−122˚−121˚

39˚

39˚

40˚

40˚

41˚

41˚

42˚

42˚

24

Elevation (km)

Coa

st R

ange

GVSN

Klamath

Clear Lake

Shasta

Lassen

Vs (km/s) Seismicitydepth (km)

SlabWindow

GV-SNasthenosphere

Gorda slab

Mantlewedge

Gordaasthenosphere

SEDGE

Fig. 9. The ‘staggered upwelling’ model illustrated by the Vs isosurfaces (3.86, 4.20, and 4.35 km/s) at 39–40°N and 41–42°N, overlain by the topography map. We suggest a hybridsource for asthenospheric upwelling near the SEDGE (dashed orange line in the topography map), first infilled by the subslab source (red arrow) after slab removal, followed by themajor flow (green arrow) from beneath the ‘GV-SN’ anomaly in its east. The colored dots indicate the location of the seismicity in the 41–42°N region. (For interpretation of thereferences to color in this figure legend, the reader is referred to the web version of this article.)

69K. Liu et al. / Earth and Planetary Science Letters 323-324 (2012) 60–71

former mantle wedge still retains a large amount of slab-derivedfluids that can suppress the solidus temperature, producing decom-pression melting as it ascends, causing surface volcanism and the sig-nificant Vs reduction under Clear Lake (Figs. 6c and 9). The ‘GV-SN’

asthenosphere provides the mantle wedge-type geochemical signa-tures (e.g., Whitlock et al., 2001). The SKS split directions are consis-tent with southeast flow from beneath Gorda, southwest flow frombeneath Cascadia, and east to west flow from the GV-SN anomaly.

a) b) c)

Fig. 10. Schematic diagrams of mantle upwelling model under the transform regime. North American plate is fixed. Bottom figures are cross-section views taken along the latitudeof Clear Lake, shown as the dark dashed lines in the top panel. (a) The MTJ was in south of the current Clear Lake location, as the Gorda plate was subducting beneath the NorthAmerican lithosphere along the Cascadia subduction zone, forming the Cascadia arc volcanoes (dark triangles). The Gorda asthenosphere was separated from that of the mantlewedge by the subducting plate. (b) An imaginary moment right as the subducting slab passes the MTJ and is replaced by asthenosphere. Northward migration of the Gordaplate creates the slab-free region. The Gorda and mantle wedge asthenospheres start to upwell and mix in the slab-free window region. (c) After mantle upwelling and mixing.Decompression melting led to the most recent surface volcanism at Clear Lake in the Coast Ranges. The abandoned mantle wedge is the “GV-SN” anomaly (red), which contributeshydrated mantle wedge asthenosphere to the melting region under Clear Lake. (For interpretation of the references to color in this figure legend, the reader is referred to the webversion of this article.)

70 K. Liu et al. / Earth and Planetary Science Letters 323-324 (2012) 60–71

This wedge-to-windowmantle flow is consistent with numerical sim-ulation based on the MTJ geometry under the transform regime (Liuand Furlong, 1992), as well as several other studies (e.g., Furlongand Govers, 1999; Furlong and Schwartz, 2004; Guzofski andFurlong, 2002). The east–west Vs cross-sections (Fig. S2) show thatthe low-velocity asthenosphere is continuous from the ‘GV-SN’anomaly to the low velocities of the slab window under the CoastRanges, consistent with this asthenospheric flow model.

This model also provides an alternative explanation for the nota-ble time lag (~3 Myr in Clear Lake) between MTJ migration and Ceno-zoic Coast Range volcanism (e.g., Dickinson, 1997). Although part ofthe delay in northern Coast Range volcanism could be caused bytime spent establishing a magmatic system (Dickinson, 1997; Liuand Furlong, 1992), or the propagation time of the flow from thearc to the window (Furlong and Schwartz, 2004), the staggered up-welling model suggests that the delay in volcanism results from thetimes of vertical and lateral transport of the former wedge into theslab window, and mixing with the Gorda asthenosphere. Until thereis sufficient accumulation of hydrated asthenosphere at shallowdepths to dramatically suppress the mantle solidus and enable largescale decompression melting, volcanism is limited to small-volumedecompression melting of Gorda asthenosphere (Henstock andLevander, 2000; Levander et al., 1998).

5. Conclusions

Overall, a variety of seismic probes identify four distinct, youngLABs and indicate mantle flow between them in the MTJ region.Joint inversion of Rayleigh wave phase velocity dispersion data(8–100 s) and Ps receiver functions provides complementary con-straints on the Vs structure at the lithospheric scale. The relativeplate motions associated with the northwestward migration of theMTJ produces three distinct modern asthenospheres: one beneaththe volcanic arc where dewatering continues, one remnant mantlewedge in which fluid flux has stopped, and the long-discussed slabwindow beneath the Coast Ranges. We suggest that the slab windowis filled by mantle material from the Gorda asthenosphere and from

the present and abandoned mantle wedge. The fourth LAB is underthe subducting Gorda plate. These four young LABs are strongly relat-ed to subduction (vertical) and northwestward (horizontal) migra-tion of the MTJ. The ‘staggered’ upwelling model, which mixesmantle flow from beneath the Gorda plate and with that from themantle wedge, is consistent with the seismic observations, and iden-tifies the source of the mantle-wedge geochemical signature seen inthe Clear Lake volcanics.

Acknowledgments

The seismic data are from the IRIS DMC. The authors would like tothank Dr. Peter M. Shearer and two anonymous reviewers for valuablecomments; R.B. Herrmann and C.J. Ammon for the CPS package; DonForsyth, and Yingjie Yang for the surface-wave code; Leland O'Driscollfor field efforts, and the PASSCAL Instrument Center. We acknowledgeMatt Fouch for sharing the SKS splitting data. K.L. acknowledges FenglinNiu and Iain Bailey for discussion on the surfacewave and joint inversioncodes, and Sally Thurner and Jen Gabler for manuscript editing. Figureswere made by GMT (Wessel and Smith, 1991) and EarthVision™. Thisstudy was funded by NSF EarthScope grant EAR-0642474 andEAR-0746379.

Appendix A. Supplementary data

Supplementary data to this article can be found online at doi:10.1016/j.epsl.2012.01.020.

References

Atwater, T., 1970. Implications of plate tectonics for the Cenozoic tectonic evolution ofwestern North America. Bull. Geol. Soc. Am. 81, 3513–3536.

Beaudoin, B.C., Godfrey, N.J., Klemperer, S.L., Lendl, C., Trehu, A.M., Henstock, T.J.,Levander, A., Holl, J.E., Meltzer, A.S., Luetgert, J.H., Mooney, W.D., 1996. Transitionfrom slab to slabless: results from the 1993 Mendocino Triple Junction seismicexperiment. Geology 24, 195–199.

Beaudoin, B.C., Hole, J.A., Klemperer, S.L., Trehu, A.M., 1998. Location of the southernedge of the Gorda slab and evidence for an adjacent asthenospheric window: re-sults from seismic profiling and gravity. J. Geophys. Res. Solid Earth 103,30101–30115.

http://dx.doi.org/10.1016/j.epsl.2012.01.020http://dx.doi.org/10.1016/j.epsl.2012.01.020

71K. Liu et al. / Earth and Planetary Science Letters 323-324 (2012) 60–71

Bensen, G.D., Ritzwoller, M.H., Barmin, M.P., Levshin, A.L., Lin, F., Moschetti, M.P.,Shapiro, N.M., Yang, Y., 2007. Processing seismic ambient noise data to obtainreliable broad-band surface wave dispersion measurements. Geophys. J. Int. 169,1239–1260.

Benz, H.M., Zandt, G., Oppenheimer, D.H., 1992. Lithospheric structure of NorthernCalifornia from teleseismic images of the upper mantle. J. Geophys. Res. SolidEarth 97, 4791–4807.

Blake, M.C., Jayko, A.S., McLaughlin, R.J., 1985. Tectonostratigraphic terranes of thenorthern Coast Ranges, California. ectonostratigraphic Terranes of the Circum-Pacific Region.

Blakely, R.J., Brocher, T.M., Wells, R.E., 2005. Subduction-zone magnetic anomalies andimplications for hydrated forearc mantle. Geology 33, 445–448.

Bostock, M.G., Hyndman, R.D., Rondenay, S., Peacock, S.M., 2002. An inverted continen-tal Moho and serpentinization of the forearc mantle. Nature 417, 536–538.

Brocher, T.M., Parsons, T., Trehu, A.M., Snelson, C.M., Fisher, M.A., 2003. Seismicevidence for widespread serpentinized forearc upper mantle along the Cascadiamargin. Geology 31, 267–270.

Cole, R.B., Basu, A.R., 1995. Nd–Sr isotopic geochemistry and tectonics of ridge subduc-tion and middle Cenozoic volcanism in western California. Geol. Soc. Am. Bull. 107,167–179.

Dickinson, W.R., 1997. Tectonic implications of Cenozoic volcanism in coastalCalifornia. Geol. Soc. Am. Bull. 109, 936–954.

Dickinson, W.R., Snyder, W.S., 1979. Geometry of subducted slabs related to SanAndreas transform. J. Geol. 87, 609–627.

Eakin, C.M., Obrebski, M., Allen, R.M., Boyarko, D.C., Brudzinski, M.R., Porritt, R., 2010.Seismic anisotropy beneath Cascadia and the Mendocino Triple Junction: interac-tion of the subducting slab with mantle flow. Earth Planet. Sci. Lett. 297, 627–632.

Forsyth, D.W., Li, A., 2005. Array analysis of two-dimensional variations in surface wavephase velocity and azimuthal anisotropy in the presence of multipathing interfer-ence. Seismic Earth: Array Analysis of Broadband Seismograms.

Fouch, M.J., West J.D., in preparation. The mantle flow field beneath the western UnitedStates.

Fulton, P.M., Saffer, D.M., 2009. Potential role of mantle-derived fluids in weakeningthe San Andreas Fault. J. Geophys. Res. Solid Earth 114.

Furlong, K.P., Govers, R., 1999. Ephemeral crustal thickening at a Triple Junction: theMendocino crustal conveyor. Geology 27, 127–130.

Furlong, K.P., Schwartz, S.Y., 2004. Influence of the Mendocino Triple Junction on thetectonics of coastal California. Annu. Rev. Earth Planet. Sci. 32, 403–433.

Furlong, K.P., Lock, J., Guzofski, C., Whitlock, J., Benz, H., 2003. The Mendocino crustalconveyor: making and breaking the California crust. Int. Geol. Rev. 45, 767–779.

Godfrey, N.J., Klemperer, S.L., 1998. Ophiolitic basement to a forearc basin and implica-tions for continental growth: the Coast Range Great Valley ophiolite, California.Tectonics 17, 558–570.

Godfrey, N.J., Beaudoin, B.C., Klemperer, S.L., 1997. Ophiolitic basement to the GreatValley forearc basin, California, from seismic and gravity data: implications forcrustal growth at the North American continental margin. Geol. Soc. Am. Bull.109, 1536–1562.

Grand, S.P., Helmberger, D.V., 1984. Upper mantle shear structure of North-America.Geophys. J. R. Astron. Soc. 76, 399–438.

Gulick, S.P.S., Meltzer, A.S., Henstock, T.J., Levander, A., 2001. Internal deformation ofthe southern Gorda plate: fragmentation of a weak plate near the MendocinoTriple Junction. Geology 29, 691–694.

Guzofski, C., Furlong, K.P., 2002. Migration of the Mendocino Triple Junction andephemeral crustal deformation: implications for California Coast Range heatflow. Geophys. Res. Lett. 29, 12-1–12-4.

Hacker, B.R., Peacock, S.M., Abers, G.A., Holloway, S.D., 2003. Subduction factory, 2, Areintermediate-depth earthquakes in subducting slabs linked to metamorphic dehy-dration reactions? J. Geophys. Res. 108 (B1), 2030. doi:10.1029/2001JB001129.

Henstock, T.J., Levander, A., 2000. Lithospheric evolution in the wake of the MendocinoTriple Junction: structure of the San Andreas Fault system at 2 Ma. Geophys. J. Int.140, 233–247.

Henstock, T.J., Levander, A., Hole, J.A., 1997. Deformation in the lower crust of the SanAndreas fault system in northern California. Science 278, 650–653.

Herrmann, R.B., Ammon, C.J., 2002. Surface waves, receiver functions and crustalstructure. Computer Program in Seismology.

Hole, J.A., Beaudoin, B.C., Henstock, T.J., 1998. Wide-angle seismic constraints on theevolution of the deep San Andreas plate boundary by Mendocino Triple Junctionmigration. Tectonics 17, 802–818.

Johnson, C.M., O'Neil, J.R., 1984. Triple junction magmatism: a geochemical study ofNeogene volcanic rocks in western California. Earth Planet. Sci. Lett. 71, 241–262.

Julia, J., Ammon, C.J., Herrmann, R.B., Correig, A.M., 2000. Joint inversion of receiverfunction and surface wave dispersion observations. Geophys. J. Int. 143, 99–112.

Kawakatsu, H., Kumar, P., Takei, Y., Shinohara, M., Kanazawa, T., Araki, E., Suyehiro, K.,2009. Seismic evidence for sharp lithosphere–asthenosphere boundaries ofoceanic plates. Science 324, 499–502.

Lachenbruch, A.H., Sass, J.H., 1980. Heat-Flow and Energetics of the San-Andreas FaultZone. J. Geophys. Res. 85, 6185–6222.

Leeman, W.P., Lewis, J.F., Evarts, R.C., Conrey, R.M., Streck, M.J., 2005. Petrologicconstraints on the thermal structure of the Cascades arc. J. Volcanol. Geotherm.Res. 140, 67–105.

Levander, A., Henstock, T.J., Meltzer, A.S., Beaudoin, B.C., Trehu, A.M., Klemperer, S.L.,1998. Fluids in the lower crust following Mendocino Triple Junction migration:active basaltic intrusion? Geology 26, 171–174.

Li, X.Q., Yuan, X.H., Kind, R., 2007. The lithosphere–asthenosphere boundary beneaththe western United States. Geophys. J. Int. 170, 700–710.

Ligorria, J.P., Ammon, C.J., 1999. Iterative deconvolution and receiver–functionestimation. Bull. Seismol. Soc. Am. 89, 1395–1400.

Liu, K.H., 2009. NA-SWS-1.1: a uniform database of teleseismic shear wave splittingmeasurements for North America. Geochem. Geophys. Geosyst. 10.

Liu, M., Furlong, K.P., 1992. Cenozoic volcanism in the California Coast Ranges—numerical-solutions. J. Geophys. Res. Solid Earth 97, 4941–4951.

Liu, K., Levander, A., Niu, F., Miller, M.S., 2011. Imaging crustal and upper mantle structurebeneath the Colorado Plateau using finite frequency Rayleigh wave tomography.Geochem. Geophys. Geosyst. 12, Q07001. doi:10.1029/2011GC003611.

Obrebski, M., Allen, R.M., Xue, M., Hung, S.H., 2010. Slab–plume interaction beneath thePacific Northwest. Geophys. Res. Lett. 37.

Piromallo, C., Becker, T.W., Funiciello, F., Faccenna, C., 2006. Three-dimensional instan-taneous mantle flow induced by subduction. Geophys. Res. Lett. 33, L08304.doi:10.1029/2005GL025390.

Porritt, R.W., Allen, R.M., Boyarko, D.C., Brudzinski, M.R., 2011. Investigation of Cascadiasegmentation with ambient noise tomography. Earth Planet. Sci. Lett. 309, 67–76.

Rondenay, S., 2009. Upper mantle imaging with array recordings of converted and scat-tered teleseismic waves. Surv. Geophys. 30, 377–405.

Schmandt, B., Humphreys, E., 2010. Complex subduction and small-scale convectionrevealed by body-wave tomography of the western United States upper mantle.Earth Planet. Sci. Lett. 297, 435–445.

Schmitt, A.K., Romer, R.L., Stimac, J.A., 2006. Geochemistry of volcanic rocks from theGeysers geothermal area, California Coast Ranges. Lithos 87, 80–103.

Shapiro, N.M., Campillo, M., Stehly, L., Ritzwoller, M.H., 2005. High-resolution surface-wave tomography from ambient seismic noise. Science 307, 1615–1618.

Snieder, R., 2004. Extracting the Green's function from the correlation of coda waves: aderivation based on stationary phase. Phys. Rev. E 69.

Thurber, C., Zhang, H.J., Brocher, T., Langenheim, V., 2009. Regional three-dimensionalseismic velocity model of the crust and uppermost mantle of northern California. J.Geophys. Res. Solid Earth 114.

Whitlock, J.S., Furlong, K.P., Lesher, C.E., Furman, T., 2001. The Juan de FucaSlab-window and Coast Range Volcanics, California: correlation between sub-ducted slab age and mantle wedge geochemistry. EOS Trans. Am. Geophys. Union82, F1186 (Abstr.).

Wilson, D.S., 1986. A kinematic model for the Gorda deformation zone as a diffusesouthern boundary of the Juan de Fuca plate. J. Geophys. Res. 91, 10 259–10 269.

Wilson, D.S., 1989. Deformation of the so-called Gorda plate. J. Geophys. Res. 94,3065–3075.

Yang, Y.J., Forsyth, D.W., 2006. Regional tomographic inversion of the amplitude andphase of Rayleigh waves with 2-D sensitivity kernels. Geophys. J. Int. 166,1148–1160.

Yang, Y., Ritzwoller, M.H., Lin, F.-C., Moschetti, M.P., Shapiro, N.M., 2008. Structure ofthe crust and uppermost mantle beneath the western United States revealed byambient noise and earthquake tomography. J. Geophys. Res. 113, B12310.doi:10.1029/2008JB005833.

Yao, H.J., van der Hilst, R.D., de Hoop, M.V., 2006. Surface-wave array tomography in SETibet from ambient seismic noise and two-station analysis—I. Phase velocity maps.Geophys. J. Int. 166, 732–744.

Zandt, G., Furlong, K.P., 1982. Evolution and thickness of the lithosphere beneathcoastal California. Geology 10, 376–381.

Zandt, G., Humphreys, E., 2008. Toroidal mantle flow through the western US slabwindow. Geology 36, 295–298.

Zhai, Y., 2010. The lithospheric structure of the Mendocino Triple Junction: implica-tions of receiver function analysis. Master thesis, Rice University.

Zucca, J.J., Fuis, G.S., Milkereit, B., Mooney, W.D., Catchings, R.D., 1986. Crustal Structureof Northeastern California. J. Geophys. Res. Solid 91, 7359–7382.

http://dx.doi.org/10.1029/2001JB001129http://dx.doi.org/10.1029/2005GL025390http://dx.doi.org/10.1029/2008JB005833

Asthenospheric flow and lithospheric evolution near the Mendocino Triple Junction1. Introduction2. FAME and USArray datasets2.1. Ballistic Rayleigh wave phase velocity (22–100s)2.2. Ambient noise dispersion data (8–30s)2.3. Ps receiver function

3. Joint inversion method4. Result and discussion4.1. Crustal heterogeneity: Franciscan Complex, ophiolite, and crustal melt4.2. Comparison with MTJSE Line 94.3. Gorda plate4.4. Slab-free window4.5. Forearc serpentinization4.6. Young asthenospheres4.6.1. Mantle wedge and Cascadia volcanoes4.6.2. ‘GV-SN’ anomaly4.6.3. SAF transform regime

4.7. Mantle flow

5. ConclusionsAcknowledgmentsAppendix A. Supplementary dataReferences